Thirty-two adult wild male and thirty-six adult wild female Monk parakeets (Psittacidae family, Psittacinae subfamily, tribe Arini) were captured inside Córdoba Zoo, Argentina (31° 25′ 31.79″ S, 64° 10′ 29.92″ W) with passive traps, following the procedure described in Valdez and Benitez-Vieyra (2016) to trap doves. Captures were made during May and June 2017, thus avoiding the molting period in the Southern hemisphere (December to April) (Navarro et al. 1992). All animals were sacrificed using sodium pentobarbital and sexed by examination of the reproductive organs; the skins were then used in the spectrophotometry measurements. We determined and compared the reflectance spectrum of twelve different body regions: forehead, crown, cheeks, nape, back, chest, belly, blue wing coverts, green wing coverts, primary and secondary remiges, and tail (on all upper surfaces). The animals were sacrificed as part of a neuroendocrine study.
Color measurements and analysis
Bird coloration cannot be accurately analyzed with tools designed for human vision, as birds perceive colors in a radically different way. Thus, we carried out all reflectance measurements within the avian spectral sensitivity range (300–700 nm, Bowmaker et al. 1997) using an Ocean Optics USB4000 spectrophotometer equipped with a halogen and a deuterium light source (830 Douglas Ave., Dunedin, FL, USA 34698), both connected to the sensor by a bifurcated fiber optic cable. Each plumage region was illuminated, and the light reflected at 45° was collected. The distance between the probe and the plumage was 4 mm, the spectrophotometer resolution 0.19 nm, the integration time 300 ms and each spectrum was the average of three readings. A white standard (Ocean Optics, WS-1-SS White Standard) was used to re-calibrate the equipment between measurements in order to correct for possible shifts in performance. Reflectance was measured using SpectraSuite software (Ocean Optics, Inc.).
After obtaining reflectance spectra, we applied a receptor-noise limited model of avian vision (Vorobyev et al. 1998) to estimate how avian receivers of chromatic signals would perceive the parakeets’ plumage colorations. This model takes into account the number and sensitivity of color receptors in the avian eye (cones) and how color information is processed in terms of signal-to-noise ratio, assuming that color discrimination is limited by photoreceptor noise. To apply an avian visual model to reflectance data, we used the pavo 2.4.0 package (Maia et al. 2013) for R (Team RC. R: A language and environment for statistical computing 2013). Cone quantum catch (Q) for each of the four avian cones was calculated under standardized daylight illumination (D65) as a representative spectrum for open habitat ambient light at midday, similar to the type of habitat that this species frequents. We used UV-type avian eyes for spectral cone sensitivities as a general representative of the parrot visual system (Bowmaker et al. 1997). Although cone parameters have not been measured in M. monachus, we used the generalized spectral cone sensitivities of the UV-type of Melopsittacus undulatus eyes since this is the only member of the Psittaciformes that has been characterized to date (Bowmaker et al. 1997; Goldsmith and Butler 2005; Lind et al. 2014). Contrasts between males–males, males‒females and females–females were characterized in units of “just noticeable differences” (JND), such that one JND represents the threshold of possible discrimination. Chromatic (dS) and achromatic (dL) distances were calculated in JNDs following the vision model using cues. Visual stimuli separated by one JND are discernible by birds, although only under ideal illumination (Olsson et al. 2015).
Morphometric measurements
For each Monk Parakeet we measured the height, width and length of the bill (from the tip to the base of the skull), length of the tarsus, total length and wing length. For this we used a digital caliper (range 0‒150 mm; resolution 0.01 mm; accuracy ± 0.02 mm) and a millimeter metal ruler (50 cm). The animals were also weighed with a PESOLA® brand spring scale (accuracy ± 2 g).
Statistical analyses
First, we examined the mean ± 2SE reflectance spectra of males and females to determine the presence of overlapping regions. Then, we calculated all pairwise chromatic and achromatic distances (measured in JNDs) among males, among females, and between males and females, for each body region. We tested whether between-sex differences were greater than within-sex differences using a permutational multivariate analysis of variance (PERMANOVA, Anderson 2001), as implemented in the adonis function of the vegan R package (Oksanen 2017). PERMANOVA is a multivariate analogue of the univariate analysis of variance (ANOVA) where multivariate variance is partitioned in the space of any arbitrary dissimilarity measure, JNDs in our case. In this test, significance is obtained by comparing the multivariate version of the F-statistic with the results obtained under a large number of permutations. We performed 24 PERMANOVAs (12 for achromatic distances and 12 for chromatic distances) to test for differences between sexes in each body part. Each test involved 999 permutations. We considered the observed distance between sexes to be significantly different from that expected by random chance if it was greater than 95% of the randomized values.
Finally, we examined the multivariate morphological variation between sexes in the six traits detailed above by principal component analysis (PCA). In addition, we applied a multivariate analysis of variance (MANOVA) to test for multivariate differences between sexes. In contrast with color cues, differences in morphology can be expressed in terms of euclidean distances among individuals, fulfilling the assumptions of common (parametric) MANOVA. When a significant effect was detected, we performed additional univariate nested ANOVAs to determine which traits accounted for the significant effect in the MANOVA.